Frontier Safety Talent: Africa

A pilot to find, screen and support overlooked African ML talent into frontier AI safety programs such as MATS and ARENA, adding new researchers to the alignment field’s talent pipeline.

Project Details

Frontier Safety Talent Africa is a 4 month pilot to identify individuals that are technically sound in African machine learning communities and help route them into frontier AI safety research pipelines.

The project will start by mapping where technical talent is concentrated across communities such as Deep Learning Indaba, Zindi, Masakhane, ILINA and university ML groups. We will then scout the high potential candidates, assess their fit and support a small cohort with orientation, reading pathways, application feedback and warm referrals to programs such as MATS, ARENA and BlueDot. The pilot will begin with an outreach around Deep Learning Indaba in August, followed by online screening and support.

The project will be led by Gideon Abako through Neuravox Foundation. Gideon brings African research networks, AI governance experience and community links through AI Safety East Africa, Effective Altruism Nigeria, Effective Altruism Kampala and EA South Africa. A part time research assistant will support mapping, outreach, candidate tracking and report writing.

The outputs will be a public African AI safety talent landscape report, a dataset of mapped communities and candidate pathways, with 10 to 15 screened candidates supported through AI safety orientation, 3 to 5 warm referrals to existing frontier AI safety programs, and a tested scouting method that can be expanded in a later phase.

Theory of Impact

Advanced AI risk reduction is constrained by the number of capable people working on technical alignment, evaluations, interpretability and other frontier safety problems. Existing safety pipelines train and mentor such people but their reach is uneven. African machine learning communities have capable technical talent, yet very few people from these communities appear to enter frontier AI safety pathways.

This pilot reduces that gap by doing targeted scouting. We will identify where ML talent sits, screen people for fit, give them a first AI safety orientation, and support the best candidates toward existing programs such as MATS, ARENA, and BlueDot. The counterfactual claim is, without this project, many of these candidates would not know these pathways exist, would not see themselves as plausible applicants or would not have the context needed to apply well.

If even a small number of high potential candidates enter technical AI safety work earlier than they otherwise would have, the project can add useful talent to a field where each capable researcher matters. The second output is the reusable map of communities, candidate profiles, and lessons from scouting that will make future African AI safety recruitment cheaper and better targeted.

People

Gideon Abako

Team Member

Track Record

Neuravox Foundation has delivered public interest technology work across AI governance, AI safety, language data infrastructure, and low-resource AI deployment. Its current work includes humanitarian AI systems for supply chains, Common voice Developer API Fund, participation in the UNDP AI Hub Language AI Accelerator, and development of community-owned language data models.

The project lead, Gideon Abako, has 8+ years of experience building data and AI systems in low-resource settings. He currently works on AI safety ecosystem growth and partnership pipelines at Lens Academy, with work spanning AI safety education organisations such as BlueDot Impact, AI Safety Quest, AISafety.com, Successif, High Impact Professionals, and 80,000 Hours. He also serves as Uganda Country Representative for AI Safety East Africa.

Gideon has led and delivered funded technical research and innovation projects, including Elrha Humanitarian Innovation Fund project on AI frameworks for supply chains, KfW funded East African Community cross-country evaluation of AI-enabled immunization stock monitoring in Uganda and Tanzania, and a Mozilla Common Voice funded language AI project. He has also built LLM provenance verification infrastructure.

Grants Received– no grants recorded

Funding Asks

grantmaking.ai Launch Round

Applied

A pilot to find, screen, and support overlooked African ML talent into frontier AI safety programs such as MATS and ARENA, adding new researchers to the alignment field’s talent pipeline.

Minimum

$18,000

Ideal

$60,000

How the money will be spent

Minimum - $18,000

The minimum budget covers the lean version of the pilot, 4 months of lead time, one in person scouting push around Deep Learning Indaba 2026, desk mapping of African ML communities, online screening of candidate leads, light application support, warm referrals to AI safety programs and a public report.

Budget breakdown:

$11,000 for lead time over four months
$4,000 for Deep Learning Indaba travel, accommodation, registration and local logistics
$1,500 for online tools, candidate tracking and communications
$1,500 for report production and publishing.

This version would test the core claim that overlooked African ML talent can be identified and routed toward frontier AI safety pathways at low cost.

Ideal - $60,000

The ideal budget funds a fuller 6 months version of the work. It adds a part time research assistant, a second scouting hub beyond Indaba, candidate preparation sessions, candidate support stipends, better tracking of referrals and a public report for future funders and AI safety programs.

Budget breakdown:

$20,000 for project lead time
$8,000 for a part time research assistant
$6,000 for scouting travel across two venues
$20,000 for candidate stipends, scholarships, and bridge support
$4,000 for the public report and a small online convening and
$2,000 for communications, tools and contingency.

The main difference between the minimum and ideal versions is candidate support. The minimum version can identify and refer people. The ideal version gives the best candidates time, guidance, and financial breathing room to prepare applications to programs like MATS, ARENA and BlueDot.

Link to budget: https://docs.google.com/spreadsheets/d/1sUkC-YC_Pcm_MKnFaiCoLzyK23McsA4rn3xG-neg92Y/edit?usp=sharing

Discussion

Private comments in this thread are only shown to approved funders and grant reviewers. Email hi@grantmaking.ai or to request access.

Grant ReviewerPrivate17h?